Integrating Models and Fusing Data in a Deep Ensemble Learning Method for Predicting Epidemic Diseases Outbreak

Abstract

Due to the continuous and growing spread of the novel corona virus (COVID-19) worldwide, it is urgent, especially in the data science era, to develop accurate data driven decision-aided methods to predict and early detect the outbreak of this epidemic disease and then to support healthcare decision makers. In this context, the main goal of this paper is to build an accurate and generic data driven method that can predict daily COVID-19 positive cases and therefore helps stakeholders to make and review their epidemic response plans. This method is based on the integration of three deep learning models: Long Short Term Memory (LSTM), Deep Neural Networks (DNN) and Convolutional Neural Networks (CNN) and takes advantage of their complementarity. The proposed method is validated on two experimental scenarios where the first one aims to validate the method on China and Tunisia case studies and the second one is based on data fusion and transfer learning process where China data and models will be reused to predict Tunisia COVID-19 outbreak. Experiment results indicate that, compared with individual learners, the stacked-DNN meta-learner, whose inputs are results of LSTM, DNN and CNN learners, achieved the best results in terms of accuracy as well as RMSE and it required the lowest time for training as well as prediction for the two scenarios. The main outcomes of this paper are i) to adopt deep learning models combined to stacking ensemble learning to accurately forecast COVID-19 positive cases and ii) to merge data and to adopt transfer learning for the prediction of confirmed cases by reusing China data, learners and meat-learners to make prediction of the epidemic trend for other countries, with less facilities of collecting data, when preventive and control measures are similar.